Absolute Quantitation of β-Lactoglobulin by Protein Liquid

Anal. Chem. 2007, 79, 5165-5172

Absolute Quantitation of β-Lactoglobulin by Protein Liquid Chromatography-Mass Spectrometry and Its Application to Different Milk Products Christoph Czerwenka,*,† Irene Maier,†,‡ Natascha Potocnik,‡ Fritz Pittner,‡ and Wolfgang Lindner*,†

Institute of Analytical Chemistry and Food Chemistry, University of Vienna, Wa¨hringer Strasse 38, 1090 Wien, Austria, and Department of Biochemistry, University of Vienna, Dr. Bohr-Gasse 9, 1030 Wien, Austria

The absolute quantitation of proteins in biological matrixes is of great interest in many fields and can be accomplished by different methodologies. Here, a method for the absolute quantitation of the whey protein β-lactoglobulin using protein liquid chromatography coupled to mass spectrometry is reported. The developed approach was characterized in detail and applied to the determination of β-lactoglobulin contents in various milk products. A special focus was placed on the recovery rates of the isolation procedure and on robust quantitation by LCMS. For these purposes protein internal standards were employed. The observed recovery rates of β-lactoglobulin from various samples ranged from 100% for whole milk to just over 50% for a strongly processed yogurt-based baby food product. The influence of processing was investigated in greater detail, showing that an increasing intensity of the applied heat treatment resulted in an increasing loss of β-lactoglobulin. LC-MS quantitation at the protein level proved to be highly suitable, avoiding a potentially problematic digestion step. The use of an appropriate internal standard to compensate for sample losses during sample workup was shown to be essential for obtaining accurate results. The accurate absolute quantitation of proteins in biological matrixes, which is of great interest and importance in clinical chemistry, pharmacology, and biomarker research, still represents a formidable analytical challenge. The two most commonly employed analytical methods for this purpose include immunoassays and liquid chromatography coupled to mass spectrometry (LC-MS). Although immunochemical methods generally offer good sensitivity and high throughput, they suffer from potential cross-reactivities. Moreover, the development of antibodies for novel targets is a tedious process. On the other hand, LC-MS combines a robust separation technique with identification and quantitation by mass spectrometry. To enable a reliable MS quantitation, a suitable internal standard (IS) must be added.1 The growing number of applications in recent years indicates that LC* To whom correspondence should be addressed. Phone: +43 1 4277 52322 (C.C.), +43 1 4277 52300 (W.L.). Fax: +43 1 4277 9523. E-mail: [email protected]. † Institute of Analytical Chemistry and Food Chemistry. ‡ Department of Biochemistry. (1) MacCoss, M. J.; Matthews, D. E. Anal. Chem. 2005, 77, 294A-302A. 10.1021/ac062367d CCC: $37.00 Published on Web 06/08/2007

© 2007 American Chemical Society

MS constitutes a highly useful methodology for the absolute quantitation of proteins in complex matrixes. The absolute quantitation of proteins by LC-MS can be carried out either at the protein level (analysis of the intact protein) or at the peptide level (analysis of signature peptides after protein digestion).2 The methodology of absolute protein quantitation by analyzing a tryptic signature peptide using an isotopically labeled synthetic analogue as the IS has been described in detail.2,3 Applications of this approach include the quantitation of plasma proteins,4,5 membrane proteins,6 and biomarkers.7-9 The advantages of the methodology include low detection limits and wide dynamic ranges.2 Moreover, the isotopically labeled peptides used as ISs can be obtained conveniently by standard peptide synthesis. At the same time, however, the IS peptides can pose several problems. First, a suitable peptide must be found whose sequence is specific for the protein.2,8 Second, the behavior of the IS peptide may differ significantly compared to that of the intact protein during the steps of the analytical workflow that take place prior to digestion. This could result in different losses of analyte and IS during sample workup and hence negatively influence the accuracy of quantitation. Another problematic issue is the tryptic digestion step. The methodology relies on the digestion of the protein being complete. However, a recent investigation showed that this may not be the case with typical digestion protocols.10 One could try to circumvent this problem by using an extended peptide precursor as the IS from which the IS peptide is released upon digestion; however, it was shown that the digestion behavior of such an extended peptide is not comparable to that of an intact protein.11 (2) Bro ¨nstrup, M. Expert Rev. Proteomics 2004, 1, 503-512. (3) Kirkpatrick, D. S.; Gerber, S. A.; Gygi, S. P. Methods 2005, 35, 265-273. (4) Cohen, A. M.; Mansour, A. A. H.; Banoub, J. H. J. Mass Spectrom. 2006, 41, 646-658. (5) Lin, S.; Shaler, T. A.; Becker, C. H. Anal. Chem. 2006, 78, 5762-5767. (6) Barnidge, D. R.; Dratz, E. A.; Martin, T.; Bonilla, L. E.; Moran, L. B.; Lindall, A. Anal. Chem. 2003, 75, 445-451. (7) Barnidge, D. R.; Goodmanson, M. K.; Klee, G. G.; Muddiman, D. C. J. Proteome Res. 2004, 3, 644-652. (8) Kuhn, E.; Wu, J.; Karl, J.; Liao, H.; Zolg, W.; Guild, B. Proteomics 2004, 4, 1175-1186. (9) Aguiar, M.; Masse, R.; Gibbs, B. F. Anal. Biochem. 2006, 354, 175-181. (10) O’Connor, G.; Burkitt, W.; Arsene, C.; Henrion, A.; Bunk, D. Presented at the 17th International Mass Spectrometry Conference, Prague, Czech Republic, August 27-September 1, 2006. (11) Barnidge, D. R.; Hall, G. D.; Stocker, J. L.; Muddiman, D. C. J. Proteome Res. 2004, 3, 658-661.

Analytical Chemistry, Vol. 79, No. 14, July 15, 2007 5165

Absolute quantitation at the protein level by LC-MS of the intact protein avoids the time-consuming and potentially problematic digestion step. Several wide-pore reversed phase columns are available nowadays for the LC separation of whole proteins, and robust electrospray ionization sources provide good ionization for unproblematic interfacing with MS. However, the availability of a suitable IS constitutes the major problem of this approach. Theoretically, the optimal IS would be an isotopically labeled analogue of the analyzed protein; however, to obtain such an IS, a laborious protein expression in an isotopically highly enriched medium is necessary. Despite the required complex procedure, this option has been employed in a LC-MS study on a protein drug candidate12 and in the MS quantitation of HPLC fractions of insulin-related proteins.13 An alternative approach consists of the use of a modified form of the target protein as the IS. Modifications that have been utilized were carbamidomethylation14 and biotinylation15 of cysteine residues and N-terminal truncation.16 In some cases it can be feasible to use a species variant of the analyzed protein (i.e., the same protein from another species) as the IS, if it has a high sequence homology with the target protein. Employing this approach, bovine heart cytochrome c was used as the IS for the MS quantitation of equine heart cytochrome c.17 A pivotal aspect in the absolute quantitation of proteins in biological matrixes is the isolation yield of the protein from the matrix and its recovery during the various sample preparation steps. Unfortunately, this topic is often neglected, and the recovery of the entire analytical process prior to the actual measurement is implicitly assumed to be 100%. If one, however, considers that other sample constituents may impede the isolation of the target protein and that proteins can quite easily undergo adsorption or degradation losses during sample workup, this negligence is not justified. A robust and accurate LC-MS quantitation becomes actually quite meaningless if only an unknown fraction of the protein present in the original sample matrix is subjected to the final measurement step. Of course, less-than-complete isolation from the matrix and losses during sample preparation can be compensated for by introducing a suitable IS. Regarding the approach of absolute protein quantitation at the peptide level, it was already discussed that an isotopically labeled analogue of the signature peptide may not constitute a suitable IS for all sample preparation steps prior to and including digestion. In practice the isotopically labeled peptide is generally added to the sample after some isolation/purification steps have already been completed2,3 or even after the digestion step,4,5 thus forgoing the chance to compensate for losses of the protein analyte during some or even all steps of the sample preparation procedure. However, in some studies the recovery of the protein analyte from the sample matrix was assessed and accounted for. A method for the absolute quantitation of myoglobin in human serum incorporated the (12) Ji, Q. C.; Rodila, R.; Gage, E. M.; El-Shourbagy, T. A. Anal. Chem. 2003, 75, 7008-7014. (13) Kippen, A. D.; Cerini, F.; Vadas, L.; Sto ¨cklin, R.; Vu, L.; Offord, R. E.; Rose, K. J. Biol. Chem. 1997, 272, 12513-12522. (14) Becher, F.; Pruvost, A.; Clement, G.; Tabet, J. C.; Ezan, E. Anal. Chem. 2006, 78, 2306-2313. (15) Bunk, D. M.; Welch, M. J. J. Am. Soc. Mass Spectrom. 1997, 8, 12471254. (16) Ji, Q. C.; Gage, E. M.; Rodila, R.; Chang, M. S.; El-Shourbagy, T. A. Rapid Commun. Mass Spectrom. 2003, 17, 794-799. (17) Gordon, E. F.; Mansoori, B. A.; Carroll, C. F.; Muddiman, D. C. J. Mass Spectrom. 1999, 34, 1055-1062.

5166

Analytical Chemistry, Vol. 79, No. 14, July 15, 2007

addition of horse myoglobin as the IS directly to the serum samples.18 In a study concerning the absolute quantitation of a rheumatoid arthritis biomarker, the recovery of the protein from serum was determined to be ∼60%.8 A comparison of the results of absolute quantitation of four serum proteins by LC-MS with reported literature values5 also showed that significant losses can occur during sample workup and need to be compensated for. In the present contribution, the absolute quantitation of the whey protein β-lactoglobulin (β-LG) in different cow’s milk products is reported. β-LG possesses a high nutritional value and interesting technological properties,19 but at the same time it constitutes one of the major allergens in cow’s milk allergy.20,21 Milk contains 3-4 g/L of β-LG,19,20 but various technological processing steps such as heat treatment or fermentation may strongly reduce this amount. The determination of the β-LG content in the various milk products was carried out by isolating the whey protein fraction, followed by LC-MS analysis of β-LG at the protein level. The recoveries of the extraction procedure from different matrixes were determined by spiking the samples with an IS prior to sample preparation. These investigations should provide detailed information on the isolation yields of β-LG and their dependence on the sample matrix. The developed methodology was applied to whole milk, constituting a rather simple sample, as well as to processed baby food products containing yogurt as a milk-related ingredient, which represented complex matrixes. EXPERIMENTAL SECTION Materials. Bovine β-lactoglobulin (mixture of variants A and B, purity >99%) was purchased from Sigma (Steinheim, Germany). Caprine and bubaline β-lactoglobulin were isolated from pasteurized goat milk (purchased at a local supermarket) and pasteurized water buffalo milk (obtained from the dairy Lat-bri, Italy), respectively, as follows: 1 L of milk was mixed with 500 mL of 0.3 M sodium chloride containing 0.2% (w/v) Triton X-100 and acidified with 6 M hydrochloric acid to pH 3-4. After incubation at 40 °C for 30 min, the solution was centrifuged at 5900g for 20 min at 4 °C and filtered. To the filtrate, ammonium sulfate was slowly added at 4 °C until a saturation level of 60% was reached after 24 h. The solution was then centrifuged at 5900g for 30 min at 4 °C. The supernatant contained the β-LG, while the other whey proteins were contained in the pellet. β-LG was then precipitated by further addition of ammonium sulfate to the supernatant at 4 °C until a saturation level of 90% was reached after 24 h, followed by centrifugation at 5900g for 30 min at 4 °C. The pellet was redissolved in 200 mL of 0.5 mM Tris/HCl pH 6.5 buffer and dialyzed against this buffer overnight. Finally, the obtained solution was lyophilized, yielding a white powder of pure β-LG. All isolation steps were monitored by SDS-PAGE. Pasteurized whole milk (3.3 g protein/100 g) was purchased at a local supermarket. Processed milk products were based on a fruit and yogurt baby food product and were manufactured in a pilot plant using a heatable stirrer vessel (batch size: 4 kg). The (18) Mayr, B. M.; Kohlbacher, O.; Reinert, K.; Sturm, M.; Gro ¨pl, C.; Lange, E.; Klein, C.; Huber, C. G. J. Proteome Res. 2006, 5, 414-421. (19) de Wit, J. N. J. Dairy Sci. 1998, 81, 597-608. (20) Wal, J.-M. Allergy 1998, 53, 1013-1022. (21) Sharma, S.; Kumar, P.; Betzel, C.; Singh, T. P. J. Chromatogr., B 2001, 756, 183-187.

Figure 1. Analysis flow chart for the determination of protein recovery, processing losses, and absolute quantitation of bovine β-lactoglobulin by LC-MS in differently processed milk products using two internal standards.

main ingredients were 45% yogurt, 12% apricot puree, and 10% banana puree. The ingredients were mixed together, heated to 90 °C, and filled into small glass jars. To the batch that was spiked with the first IS prior to further processing, 3.06 g of caprine β-LG ()765 µg/g product) was added after heating to 90 °C (Figure 1 left-hand side). Subsequent processing was performed in the form of a heat treatment in the following three variants: The glass jars were either cooled down immediately below 30 °C with cold water (denoted as “no heat treatment”), incubated at 95 °C in a spray autoclave for 26 min before being cooled down immediately (“medium heat treatment”), or incubated at 110 °C in a spray autoclave for 26 min before being cooled down immediately (“strong heat treatment”). Sample Preparation: Isolation of Whey Proteins from Milk Products. Sample preparation consisted of lipid removal and casein precipitation to isolate the whey proteins. For whole milk, a simple and rapid procedure that combined both steps22 was used: After an exactly weighed 1 mL aliquot of milk was spiked with the first IS by adding 1 mg of caprine β-LG, 100 µL of 5% (v/v) acetic acid and 200 µL of dichloromethane were added. The mixture was vortexed briefly and centrifuged for 10 min at 3480g, yielding three layers with the top one constituting the whey protein isolate. For the processed milk products, 20 g of each sample was mixed with the same amount (w/w) of a 0.3 M sodium chloride solution containing 0.2% (w/v) Triton X-100. Samples from the previously unspiked batch were spiked with the first IS by adding 20 mg of caprine β-LG ()999 µg/g product) at this stage (Figure (22) Chen, R.-K.; Chang, L.-W.; Chung, Y.-Y.; Lee, M.-H.; Ling, Y.-C. Rapid Commun. Mass Spectrom. 2004, 18, 1167-1171.

1 right-hand side). After extensive homogenization by shaking with glass beads for 1-2 h at room temperature, the mixture was centrifuged at 30 360g for 45 min at 4 °C. The aqueous layer was filtered, acidified with 6 M hydrochloric acid to pH 2, and incubated at 37 °C for 20 min. Centrifugation at 30 360g for 20 min at 4 °C, followed by filtration, yielded the whey protein isolates. The amounts of obtained whey protein isolates were determined volumetrically. The sample preparation procedure was carried out in duplicate for each sample. Prior to LC-MS analysis, the samples were diluted appropriately with water, and an aqueous solution of the second IS (bubaline β-LG) was added at a final concentration of 125 µg/mL. Liquid Chromatography-Mass Spectrometry. LC-MS analysis of the whey protein isolates was performed on a system consisting of a HP 1100 series HPLC instrument comprising a binary pump, autosampler, and column oven (Agilent Technologies, Waldbronn, Germany) coupled via a splitting-T (split ratio ∼1:100) to a PE Sciex API365 triple-quadrupole mass spectrometer (MDS Sciex, Concord, Canada) equipped with an electrospray ion source. The LC separation was carried out on a Supelco Discovery Bio Wide Pore C8 column (150 mm × 2.1 mm, 3 µm) from SigmaAldrich (Vienna, Austria). Elution was performed at a flow rate of 0.25 mL/min with water containing 0.5% (v/v) acetic acid as eluent A and acetonitrile containing 0.5% (v/v) acetic acid as eluent B, employing a linear gradient from 35% B to 50% B in 16 min. The column was thermostated at 40 °C. The injection volume was 10 µL. MS analysis was carried out in the scan mode with positive ionization using an ionspray voltage of 4200 V and scanning a m/z range of 800-2200. From the total ion chromatograms, summed spectra were generated incorporating all different β-lactoglobulins. The obtained spectra were deconvoluted using Mag-Tran 1.02 software, which is based on a charge state deconvolution algorithm developed by Zhang and Marshall.23 For peak finding, a signal-to-noise threshold of 2 and a mass accuracy of 0.3 Da were set. The intensities at the respective peak maxima in the deconvoluted spectra were used for quantitation. To construct the calibration curves for bovine and caprine β-LG, six multi-β-LG standards were employed, which contained 25/12.5, 50/25, 100/50, 250/125, 500/250, and 1000/500 µg/mL of bovine/ caprine β-LG, respectively, as well as 125 µg/mL bubaline β-LG. All LC-MS measurements were performed in triplicate. RESULTS AND DISCUSSION Analytical Method for the Absolute Quantitation of β-Lactoglobulin in Milk Products by Protein LC-MS. The method for the absolute quantitation of β-LG in cow’s milk products consisted of a sample preparation procedure to remove lipids and caseins, which was followed by a LC-MS analysis of the whey protein isolate (Figure 1). Two internal standards were employed to determine on the one hand the recovery of the isolation procedure and β-LG losses effected by different processing and to enable accurate MS quantitation on the other hand. These internal standards are also referred to as the first and the second IS, respectively. For an application of the method in routine analysis, the use of one IS which is added to the milk product before beginning the sample preparation procedure would of (23) Zhang, Z.; Marshall, A. G. J. Am. Soc. Mass Spectrom. 1998, 9, 225-233.

Analytical Chemistry, Vol. 79, No. 14, July 15, 2007

5167

Figure 2. Comparison of the protein sequences and some properties of the analyte bovine β-lactoglobulin (variants A and B) and the internal standards used for recovery and determination of processing losses (caprine β-lactoglobulin) and for MS quantitation (bubaline β-lactoglobulin), respectively. Amino acid differences between species are marked in bold, while those between variants A and B of bovine β-lactoglobulin are marked in bold italic. Protein properties were calculated with ProtParam on expasy.org (http://expasy.org/doc/expasy_tools05.pdf, accessed December 7, 2006).

course be sufficient, as this approach would allow the compensation (but not the determination) of analyte losses during protein isolation and of variations of the MS measurements. However, in the current study one aim was to establish the recoveries of β-LG from different matrixes. As no suitable blank matrix (milk product being completely devoid of β-LG), which could have been spiked with a known amount of the analyte for the determination of sample preparation recovery, was obtainable, the first IS was introduced into the (processed) milk product before starting with the isolation procedure (“spiking before sample preparation”, see Figure 1 right-hand side). From the samples that were spiked at this stage, the recoveries of the isolation procedure could be determined from the concentrations of the first IS measured by LC-MS. For the yogurt-containing baby food products that underwent different processing (varying degrees of heat treatment), the same IS was also introduced into a separate sample batch at the stage of the initial milk product prior to the processing (“spiking before processing”, see Figure 1 left-hand side). In these samples, the first IS experienced both processing and sample preparation losses until it was quantitated by LC-MS. Using the recoveries determined from the samples spiked before sample preparation, the losses of β-LG caused by the different processing conditions could be calculated. The whey protein isolates were subjected to LC-MS analysis without further purification, as the second major whey protein besides β-LG, R-lactalbumin, was cleanly separated from β-LG in the course of the LC separation (see below). The chromatographic separation and MS detection took place at the level of the intact protein, thus avoiding a time-consuming and potentially problematic digestion step (see above). A constant amount of the second IS was added to the whey protein isolates before the samples were injected into the LC-MS system. This second IS was used to 5168

Analytical Chemistry, Vol. 79, No. 14, July 15, 2007

correct for variations in the MS response to enable robust and accurate quantitation. The mass spectrometer was operated in the full scan mode in order to acquire the entire charge state distributions (CSDs) of the proteins. This avoided any potential reproducibility and accuracy problems that could occur in selectedion or selected reaction monitoring (SRM) modes upon shifts in the CSD. The drawback of the full scan mode, its lower sensitivity, was deemed not to be a problematic issue in the present case, as the concentrations of β-LG in milk are high (in the mg/mL range). Thus, even if only a few percent of the original concentration should remain after the processing and sample preparation procedures, they are picked up easily. From the total ion chromatograms, spectra were generated that contained the signals of the analyte and both internal standards. After deconvolution, the intensity ratios of analyte/second IS and first IS/second IS were used to determine the concentrations of the analyte and the first IS, respectively. It is noted that the presented method only addresses the intact (whole) β-LG protein and does not include fragments or derived peptides that may still have immunological activity.24 The protein analyte bovine β-LG that was quantitated in the cow’s milk products occurs in two main natural variants, which are called bovine β-LG A and bovine β-LG B.25 The variants differ by two amino acid substitutions at positions 64 and 118, which result in a mass difference of 86 Da (Figure 2). The two variants were always summed up in the present study to give the (total) amounts of bovine β-LG. Species variants of bovine β-LG that show (24) Se´lo, I.; Ne´groni, L.; Cre´minon, C.; Yvon, M.; Peltre, G.; Wal, J.-M. Int. Arch. Allergy Immunol. 1998, 117, 20-28. (25) Ng-Kwai-Hang, K. F.; Grosclaude, F. In Advanced Dairy Chemistry, 3rd ed.; Fox, P. F., McSweeney, P. L. H., Eds.; Kluwer Academic/Plenum Publishers: New York, 2003; pp 739-816.

Figure 3. Overlaid extracted ion chromatograms of the 15+ charge states of a multi-β-lactoglobulin standard containing 1000 µg/mL bovine β-lactoglobulin (variant A, dark blue trace; variant B, light blue trace), 500 µg/mL caprine β-lactoglobulin (green trace), and 125 µg/ mL of the MS internal standard bubaline β-lactoglobulin (red trace).

a high sequence homology were chosen as the two internal standards. Caprine β-LG from goat (Capra hircus) and bubaline β-LG from water buffalo (Bubalus bubalis), which were isolated from the respective milks, were used as the first and the second IS, respectively. Their sequences are aligned with those of the two variants of bovine β-LG in Figure 2. Both caprine and bubaline β-LG show greater homology with variant B of bovine β-LG, from which they differ in six and two amino acid residues, respectively. As the properties of an IS should match those of the analyte as closely as possible, some characteristics (molecular weight, pI, aliphatic index, and average hydrophobicity) of bovine β-LG A and B, caprine β-LG, and bubaline β-LG were calculated and are compared in Figure 2. It can be seen that the values of these characteristics of both caprine and bubaline β-LG match those of bovine β-LG A and B very closely. Thus, caprine and bubaline β-LG should exhibit practically identical behaviors as that of bovine β-LG during the entire analysis, hence being excellently suited as internal standards. To further substantiate the good suitability of caprine and bubaline β-LG as internal standards, their chromatographic and mass spectrometric behaviors were evaluated and compared with those of bovine β-LG. The LC retentions of the analyte and the two internal standards are shown in Figure 3. It is noted that the two variants of bovine β-LG are clearly distinguishable on the chromatographic time scale, although they are not baseline resolved. Bubaline β-LG coelutes entirely with bovine β-LG B, whereas caprine β-LG exhibits a minimally lower retention. The position of bubaline β-LG between caprine and bovine β-LG on the chromatographic time scale results in it being well suited to serve as the IS for the MS quantitation for both proteins. The original and deconvoluted mass spectra of the various β-lactoglobulins are shown in Figure 4. The CSDs of all three β-lactoglobulins are equivalent, showing a range of charge states from 9+ to 18+ with a maximum at 15+ (Figure 4a,c,e). In the spectrum of bovine β-LG, each charge state consists of a “doublet”, which represents the variants A and B (Figure 4a inset). Deconvolution of the mass spectra yielded the molecular masses

of the proteins, which were in good agreement with the theoretical values (Figure 4b,d,f). Thus, the analytical behaviors of caprine and bubaline β-LG excellently mimicked that of bovine β-LG, confirming the suitability of the proteins chosen as internal standards. Multi-β-LG standards containing bovine β-LG, caprine β-LG, and bubaline β-LG were analyzed by LC-MS to construct the calibration curves for bovine β-LG (analyte) and caprine β-LG (first IS for determination of recovery and processing losses). The calibration curves displayed very good linearity over a range of 25-1000 µg/mL for bovine β-LG and 12.5-500 µg/mL for caprine β-LG (Figure 5). Analysis of Whole Milk. The analysis of whole cow’s milk, which constitutes the most simple milk product and which has experienced practically no processing (except homogenization and a brief, mild heat treatment in the form of pasteurization), proved to be straightforward. It was possible to isolate the whey protein fraction in a single-step procedure that simultaneously removed lipids and caseins. LC-MS analysis confirmed that no further purification was necessary. The employed reversed-phase LC conditions resulted in the elution of the milk sugar lactose and other highly polar components (e.g., salts) with the void, while the whey proteins eluted within less than 15 min (Figure 6). The β-lactoglobulins (the analyte bovine β-LG and the internal standards caprine β-LG and bubaline β-LG) were cleanly separated from the second major whey protein in cow’s milk, bovine R-lactalbumin. The results of the LC-MS quantitation are presented in Table 1. Replicate analysis of the whey protein isolates proved the robustness of the LC-MS method, with the relative standard deviations of the concentrations of the analyte bovine β-LG and the first IS caprine β-LG being generally below 10% (n ) 3). The results for the two separately processed sample aliquots were also in good agreement, indicating acceptable reproducibility of the isolation procedure. The recovery rates of β-LG which were determined by comparing the concentrations of the IS caprine β-LG that were found by LC-MS with those present in the original milk sample after spiking showed that a complete recovery was achieved with the rapid one-step sample preparation method employed. The concentration of bovine β-LG in the analyzed whole milk was determined to be 3.25 ( 0.15 g/L (n ) 6), which is in excellent agreement with previous literature reports.19,20 Analysis of Processed Milk Products (Yogurt-Based Baby Food Products). The set of analyzed processed milk products was based on a baby food product that contained yogurt and two fruit purees as the main ingredients (no other milk-derived ingredients). The mixture was heated to 90 °C and filled into glass jars. This stage constituted the unprocessed milk product. Subsequent processing consisted of different heat treatments (none/medium/strong) giving the various processed milk products. The more complex composition of these food products and their extensive processing, which is expected to effect pronounced reactions of and between the different components such as Maillard reactions, decompositions, and aggregations, resulted in the processed milk products constituting highly complex matrixes, from which β-LG had to be isolated. It was found that the simple one-step procedure that was suitable for the isolation of β-LG from Analytical Chemistry, Vol. 79, No. 14, July 15, 2007

5169

Figure 4. (a, c, e) ESI mass spectra and (b, d, f) deconvoluted mass spectra of (a, b) bovine β-lactoglobulin, (c, d) caprine β-lactoglobulin, and (e, f) bubaline β-lactoglobulin.

whole milk did not work efficiently for the processed milk products. Therefore, a more elaborate procedure was developed with delipidation and casein precipitation being carried out in two separate steps. The LC-MS analysis of the whey protein isolates proceeded in the same way as for the whole milk samples. No interfering signals were observed in the elution window of the whey proteins. In analogy to the whole milk samples, the recovery rates of the β-LG isolation from the processed milk products were determined from samples that had been spiked with the first IS caprine β-LG prior to starting the sample preparation procedure 5170

Analytical Chemistry, Vol. 79, No. 14, July 15, 2007

(Figure 1 right-hand side). The obtained results are given in Table 2. The reproducibilities between replicate LC-MS measurements of the same whey protein isolate and between the two isolates of the same processed milk sample were again rather decent, with relative standard deviations generally below 20%. The recovery rates showed an interesting picture for the differently processed milk products. While over 80% of the β-LG was recovered from the milk product that had undergone no heat treatment, the isolation yield was only approximately 60% for the product that had experienced medium heat treatment, with an even lower yield (54%) for the strongly heated product. Concurrent with the

Figure 5. Calibration curves of (a) bovine and (b) caprine β-lactoglobulin using bubaline β-lactoglobulin as the internal standard for MS quantitation. Errors bars indicate standard deviations.

Figure 6. Total ion chromatogram of a whey protein isolate from whole milk spiked with caprine β-lactoglobulin for the determination of the recovery of the isolation procedure and with bubaline β-lactoglobulin for MS quantitation.

decrease of recovery rates, the volumes of the whey isolates decreased with increasing intensity of the applied heat treatment. These findings clearly show that a complete recovery of a protein analyte from a complex matrix should not be taken for granted and that technological processing may influence the matrix which in turn can affect the isolation yield of the protein. Thus, for quantitation methods of proteins in biological matrixes, losses during sample workup are likely and need to be accounted for. Moreover, it has to be considered that even seemingly similar matrixes can lead to significantly different protein recoveries. Although traces of bovine β-LG were detected in most of the processed milk products, the signals were in all cases far below the lowest calibration standard (25 µg/mL ) 1.4 pmol/µL) and not suitable for quantitative evaluation. Taking the lowest calibra-

tion standard as a benchmark and correcting for the different recovery rates, the concentrations of bovine β-LG in the processed milk products are in all cases significantly below 24-46 µg/g. If the yogurt part of the food product were to be replaced with milk, a β-LG concentration of approximately 1410 µg/g food product would be expected (calculated from 45% milk-related ingredient and 3.13 mg β-LG/g milk). From this comparison, a diminishment of β-LG of more than 96.7% between the native whole milk as the precursor of the yogurt ingredient and the yogurt as present in the final food product can be derived. This greatly decreased β-LG content can be attributed to two potential causes: The fermentation of milk to produce yogurt might have considerably reduced the amount of β-LG by degradation processes. On the other hand, the heating of the product mixture could also have effected a pronounced decrease of the β-LG content. The investigation of the effect of different heat treatments of the milk product on its β-LG content should elucidate this issue. A separate batch of the nonprocessed yogurt-based baby food product was spiked with the first IS caprine β-LG prior to the subsequent different heat treatments (Figure 1 left-hand side). The samples that were obtained from this batch were processed and analyzed without any further addition of the first IS. The influence of different processing conditions on β-LG was then derived from the final concentrations of caprine β-LG determined by LC-MS after correcting for the isolation yields (see above). The results are shown in Table 3. The samples that had experienced no heat treatment exhibited a loss of β-LG of around 18%, while those that had undergone a heat treatment of medium intensity displayed losses of about 90%. Finally, in the strongly heat treated samples no caprine β-LG was detected anymore, indicating a complete loss of β-LG. Thus, the extent of the diminishment of β-LG during processing is clearly correlated with the intensity of the heat treatment. Interestingly, also the samples that had undergone no (final) processing step showed some loss of β-LG. These samples were produced by immediately cooling down the samples after the ingredients had been mixed, the mixture heated to 90 °C and filled into jars. The IS was added immediately before the filling step. The short period of time during the cooling process where the samples were still at elevated temperatures is thought to be responsible for the observed decrease in β-LG concentration, which is low compared to the samples that experienced a (final) heat treatment before being cooled down. As both the medium and strongly heat treated samples were incubated for the same period of time but at different temperatures, it can be deduced that the diminishment of β-LG caused by heat treatment increases with increasing values of the applied temperature. Finally, the results from the heat treatment investigation, showing extensive losses of β-LG of 90% and more, indicate that the reason for the large decrease of the β-LG concentration in the yogurt fraction of the food product compared to native milk can be attributed to a large extent to thermal treatments. These occur in the form of a heating step during the production of the initial unprocessed milk product (Figure 1 topmost stage) from yogurt and other ingredients as well as during the production of the yogurt itself, where milk is heated prior to starting the fermentation. Pure yogurt already showed markedly lower β-LG concentrations than whole milk (data not shown). The fermentation process does not seem to play a major role in the Analytical Chemistry, Vol. 79, No. 14, July 15, 2007

5171

Table 1. Recovery Rate and Concentration of β-Lactoglobulin Determined in Whole Milk by LC-MS Using Caprine β-LG as IS for Sample Preparation and Bubaline β-LG as IS for MS Quantitation

sample

bovine β-LGa µg/mL

caprine β-LGa µg/mL

whey isolate µL

recovery rate β-LG isolation %

whole milk/1c whole milk/2d

3977 (2.0) 4374 (3.0)

1335 (9.3) 1271 (12.8)

825 830

105.0 107.2

concentration of bovine β-LG in milkb mg/g

concentration of bovine β-LG in milkb mg/mL

3.02 3.23

3.13 3.36

a Values given are means from triplicate LC-MS measurements (RSDs in % in parentheses). b Values are corrected for the respective recovery rates. c 1.0335 g of milk spiked with 1.049 mg of caprine β-LG. d 1.0414 g of milk spiked with 0.984 mg of caprine β-LG.

Table 2. Recovery Rates of β-Lactoglobulin Isolation from Processed Milk Products heat treatment

caprine β-LGa µg/mL

whey isolateb mL

caprine β-LG µg/g

recovery rate β-LG isolationc %

none/1 none/2 medium/1 medium/2 strong/1 strong/2

601 (15.3) 594 (17.8) 534 (16.3) 662 (22.3) 549 (11.3) 535 (3.6)

28 28 21 21 20 20

841 831 560 694 549 534

84.2 83.2 56.0 69.5 54.9 53.5

a Values given are means from triplicate LC-MS measurements (RSDs in % in parentheses). b Obtained from 20 g sample. c Each sample was spiked with 999 µg/g caprine β-LG prior to β-LG isolation.

Table 3. Losses of β-Lactoglobulin upon Different Processing

heat treatment none/1 none/2 medium/1 medium/2 strong/1 strong/2

β-LGa

caprine µg/mL

392 (3.6) 383 (19.0) 100 (11.0) 77 (5.2)